Knowledge Engineering and Knowledge Management

Galway City, Ireland, 8-12 October 2012.

Event website.

By Laura Moss, Agent and Clinical physicist, NHS Greater Glasgow & Clyde and the University of Aberdeen.

Highlights

The highlight of the conference was hearing about how semantic web applications are being developed in the 'real world' and the issues that these researchers are encountering.

Conference report

The 18th International Conference on Knowledge Engineering and Knowledge Management (EKAW 2012) was held in Galway, Ireland, and was attended by over 120 academics, entrepreneurs, and industrial researchers interested in representing and managing knowledge in a computerized form. Organisers of the conference came from the locally based DERI (http://www.deri.ie/) research institute which is a leading web science research institute. Currently, researchers in the field are working on developing tools and projects supporting the Semantic Web, in which data is linked and represented using semantic web technologies, enabling computers to do more useful work.

The main conference was spread over three days and featured a total of 30 presentations. In addition, there were two days of workshops prior to the conference and a lively poster and demo session. Topics covered by the conference included: natural language processing, the extraction and representation of knowledge, ontology engineering and evaluation, and linked open data. A clear theme, reflected in the keynote talks, was the impact of knowledge engineering and management research from the past 25 years (since the conference series began) on current, real-world applications. There was a feeling during the meeting that this research community had reached a point where these technologies are now starting to become increasingly used out with the research community, for example by the business community. Prospects for the future are also exciting, as knowledge is playing an important role on the Web; specifically the increasing availability of information in the form of linked open data (http://linkeddata.org/) which allows information to be combined in interesting ways leading to the development of powerful applications.

The three keynote talks all discussed the development of knowledge (or semantic) driven real world applications. On the first day, the keynote presentation was given by Prof. Martin Hepp from the Universität der Bundeswehr München in Germany. His presentation was titled ‘From Ontologies to Web Ontologies: Lessons Learned from Conceptual Modelling for the World Wide Web’. In computer science, ontologies can be considered as a formal specification of concepts in a domain and their relationships and provide structural frameworks for organising information. In his talk Prof. Martin Hepp suggested that there was a fundamental difference between the traditional, philosophical variant of an ontology compared to the type of ontologies that are required for use on the web, i.e. web ontologies. He described some of the challenges faced when developing data structures that are suitable for use on the Web, for example, data quality issues, the processing of huge amounts of data, and the limitations of real world deployment environments such as the skills of employees and pre-existing corporate policies. He argued that ‘web ontologies’ need to work for a broad user base, contain a minimal number of conceptual elements, should provide HTML documentation, should be understandable and useable for web presence, and minimise namespace traversals. Overall, it was proposed that ontology engineering needed to become more of an empirical science that should be grounded in the domain, rather than being dominated by logical representations.

On the second day Dr. Michael Uschold gave a talk on his experiences building real-world enterprise ontologies for large organisations, in particular his experiences building an enterprise ontology for Sentara Healthcare. He suggested that the starting point for ontology engineering should be to identify the scope for the ontology that you are building e.g. identify the key business areas. This stage should involve interviews with employees from all levels and areas of the workplace and shouldn’t be swayed by any preconceptions that you may have of the organisation. Interestingly, this point reflected some points the previous keynote speaker made about making sure that you spend time understanding the domain. Other advice that ontology engineers may find useful included: to build ontologies that have a small number of high level disjoints and have very few orphan categories; use as few primitives as possible (making the ontology easier to learn and reuse); and to try and apply the concept of ‘layered ambiguity’, in which low level details are hidden and expanded when actually needed. Problems encountered during the development of the enterprise ontology included: evaluation issues e.g. how to ensure the ontology is correct, and knowing what to show the client. In particular for the later problem, he proposed that tools are required that enable domain users to view the enterprise ontology, specifically the tools should be: intuitive; ‘layout friendly’; easily distributable; inference aware; persistent; and dynamic (i.e. detail is shown upon demand), and actively engage the client.

On the third day, Dr Lee Harland presented an overview of the EU funded Open PHACTS Project (http://www.openphacts.org/). In his talk he described some of the problems faced by the pharmaceutical industry when developing new drugs e.g. it costs billions of dollars to develop a drug and many drugs still fail during clinical trials. The aim of the Open PHACTS project is to develop a semantic technology platform that allows knowledge & data to be integrated from numerous sources, for example internal information available in different drug development companies and open public data. The project will use ontologies, shared vocabularies, and data will be expressed as Linked Open Data. It is hoped that by the end of the project, there will be an open service that enables scientists interested in developing new drugs to query this large knowledge source in a method that still allows for competition between commercial companies. The project is due to be completed in 2014.

A number of the papers presented during the conference & associated workshops were largely theoretical; however, some tools which semantic software developers may find interesting included: OOPS, which helps developers to automatically detect errors in ontologies; SlideWiki which is a crowd-sourcing platform for the elicitation and sharing of corporate knowledge using presentations; and RightField a Java application that allows ontology annotation support for scientific data in Microsoft Excel or Open Office spreadsheets. Other software tools mentioned during the talks included the following tools to support collaborative ontology building, OntoWiki (http://ontowiki.net/Projects/OntoWiki), Moki (https://moki.fbk.eu/website/index.php) and SMW (http://sourceforge.net/projects/smwontoeditor/); and tools to help consistency during the building of knowledge bases, e.g. INSIGHT and LODE (http://sourceforge.net/projects/lode/), software which automatically produces human readable documentation of ontologies.

To help promote the Software Sustainability Institute, I displayed an information slide at the end of two presentations that I gave on the capture and refinement of medical knowledge during the KMED workshop. Additionally, I presented a poster during the main EKAW conference and placed Software Sustainability Institute leaflets next to the poster for people to take. Several people enquired about the work the institute does; however, they were largely academics from non-UK institutions.